Forecasting Cascading Effects in Network Models as Applied to Urban Services Provision Assessment

This paper proposes a method for evaluating the urban services provision based on a block-network city model in the context of cascade effects forecasting. Cascade effects are effects that arise in the urban environment as a result of the local territory development. Such effects can affect unknown quantity of city residents, which must be forecasted. The proposed method combines two approaches to assessing provision: predicting the number of interactions and evaluating provision according to regulatory requirements, as well as assessing the diversity of non-standardized urban services. The approach based on reducing the predicting of interactions to the knapsack problem using the GraphSage graph neural network architecture showed the best performance in evaluating provision with standardized services. Meanwhile, for assessing diversity, an approach based on block-by-block evaluation of the Shannon entropy index was used, followed by the use of weighting coefficients depending on the distance between city blocks.